Compress a PDF without wrecking its quality

"Compress this PDF" is rarely the real request. The real request is usually "make this fit under a 10 MB limit without making it unreadable". Those are different problems, and knowing where the bytes are is what keeps quality intact.

What makes a PDF large

Open almost any oversized PDF and the size breaks down roughly like this:

Contributor	Typical share	Compressible?
Embedded raster images and scans	Very large	Yes, with care
Embedded fonts	Moderate	Only by subsetting
Page text and vector drawings	Small	Barely worth it
Metadata, unused objects, duplicate resources	Small to moderate	Yes, losslessly

A 40-page text report is normally a few hundred kilobytes. A 4-page scanned contract can be 20 MB. If your file is large, look for images first.

Do the lossless work first

Before any quality is sacrificed, several reductions cost nothing visually:

• Removing objects that are no longer referenced by any page.
• Deduplicating resources that were embedded once per page instead of once per document.
• Recompressing the document's object streams with a more efficient scheme.
• Dropping metadata and revision leftovers you do not need to keep.

On PDFs assembled by exporting, merging, and re-exporting repeatedly, this structural cleanup alone can be a substantial reduction. The compress PDF workspace reports which contributors dominate your specific file, which is more useful than a generic quality slider.

When downsampling images is justified

Downsampling reduces the stored pixel dimensions of an image. It is justified when the stored resolution is far beyond what the page actually uses. An image placed at 5 cm wide but stored at 4000 pixels across is being displayed at roughly 2000 DPI, when 300 DPI is already print quality and 150 DPI is comfortable for screen reading.

Reasonable targets:

• 300 DPI when the document will be printed or contains fine print.
• 200 DPI for scanned text that will mostly be read on screen.
• 150 DPI for reference copies and internal circulation.

Below roughly 150 DPI, small text in scans starts to break down, and any optical character recognition you run later becomes noticeably less accurate.

Protect the searchable layer

A common shortcut is to convert every page to a JPEG and rebuild the PDF from those images. The file gets smaller, but the document stops being searchable, stops being selectable, and becomes hostile to screen readers.

If a document already has a text layer, keep it. If it does not, consider adding one with optical character recognition before compressing, so the smaller file is also more useful than the original.

The reverse workflow is occasionally correct: when you genuinely need images, export the pages deliberately with the PDF to image converter rather than letting a compressor silently flatten the document.

Verify before you send

Compression is only successful if the result still does its job. Check:

• The smallest text on the busiest page, at 100% zoom and at print size.
• Tables, stamps, and signatures, which show compression artefacts earliest.
• Photographs with smooth gradients, where banding becomes visible quickly.
• The page count and page order, which should be unchanged.

Keep the original. Compression is lossy by definition once images are resampled, and there is no way back from a discarded pixel.

When the file is still too large

If a document remains oversized after sensible compression, the problem is usually scope rather than encoding. Split out the pages the recipient actually needs, or scan at a sensible resolution in the first place with a document scanning workflow instead of photographing pages at maximum camera resolution.